Research On Algorithms And Applications Related To H.264/AVC Residual Coding

With the development of communication and computer technology, source coding has been given more and more advanced requirements in practical application performances. H.264 is the most recent video coding international standard, employing many new coding technologies. This standard can achieve two times the coding efficiency compared with former standard with the same image quality. Therefore, it is considered the most influential video coding standard.Residual coding processes the data after prediction and writes them to the code stream, achieving further compression of video sequences. It is an important part of video coding technology. Since residual coding generates the kernel of the code stream, the capability of its relative algorithms has tremendous impacts the performance of the encoder and decoder. The research work of this thesis is mainly on optimized algorithms of residual coding.Based on the analysis of the performance of existing residual coding algorithms, the following tasks have been carried out in this thesis.Firstly, A residual coding algorithm employing compressed sensing (CS) theory is proposed. CS theory explores the inherent correlation of sparse signals for more efficient signal sampling, compression and recovery. Our algorithm incorporates the CS theory into the traditional DCT based coding method to achieve better compression efficiency.Secondly, A Multiple Description Coding algorithm employing CS theory is proposed. The CS theory states that a signal can be precisely reconstructed from only a small set of measurements. Thus our algorithm generates multiple descriptions in the frequency domain, each description able to recover the original data, and the more descriptions the better quality. Experiments have demonstrated the feasibility and advantage of the proposed MDC scheme.Finally, The existing adaptive quantization matrix selection (AQMS) algorithm is analyzed in depth, and a fast algorithm is proposed, maintaining AQMS at macroblock level while eliminating iterative matrix parameter calculation at frame level. Therefore, computational complexity is greatly reduced while keeping the same coding gain.